Every hosting provider advertises a number with a decimal point and a few nines. Here is the math behind it, what each tier really permits, and why your measured uptime and your SLA uptime are rarely the same number.
The formula
Uptime percentage = (total time - downtime) / total time × 100. For a 30-day month (43,200 minutes), 43 minutes down gives you 99.9%. The arithmetic is simple; what counts as downtime is where the arguments start.
What the nines allow per month
- 99% — 7.2 hours. Two nines is a hobby, not a service level.
- 99.9% — 43 minutes. One bad deploy or a single provider hiccup can eat the whole month.
- 99.95% — 22 minutes. The tier where redundancy starts being mandatory.
- 99.99% — 4.3 minutes. Essentially zero tolerance for single-component failures.
- 99.999% — 26 seconds. Five nines is an architecture, not a promise.
Run the numbers for your own numbers with the free uptime calculator — it converts between downtime minutes and percentage in both directions.
Why your number and the SLA number differ
Providers calculate SLA against their definition of downtime: usually total unreachability of the service, measured from their edge, excluding scheduled maintenance windows they announce in advance. Partial degradation — the API works but is ten times slower — often does not count, even though your users experience it as an outage.
Your measured uptime should be stricter than your SLA. Measure from where your users actually are, count degraded responses as partial downtime, and do not exclude anything. The number you get from independent external monitoring is the one to optimize; the SLA number is the one to invoice against when they miss it.